Skip to main content
@mcp-b/smart-dom-reader is a DOM extraction library for agent workflows. It turns a live document into structured data, interactive element maps, and stable selectors while keeping token usage low.

​
When to use this package

  • You are building DOM-aware tools for agents or browser automation.
  • You need structured page summaries or ranked selectors.
  • You want DOM extraction without bringing in a full MCP runtime.

​
When not to use this package

  • You only need to expose browser-page tools. Use the runtime packages instead.
  • You need Chrome extension API tools rather than DOM extraction. Use @mcp-b/extension-tools.

​
Where it sits in the package graph

This package sits in the tooling layer and can be used standalone. Other tooling packages, including @mcp-b/extension-tools, can build on top of it.

​
First step

Choose whether you need one-shot extraction or staged extraction. The reference page documents both approaches, selector outputs, and the optional MCP integration.

Reference

Extraction APIs, selector ranking, and result shapes.

@mcp-b/extension-tools overview

Tool classes that build on smart-dom-reader in extension contexts.

@mcp-b/chrome-devtools-mcp overview

Tooling path when you want DevTools-based automation instead.

Tool design

Design guidance for tools that expose DOM-derived capabilities.